Implementation of Text Mining in Predicting Consumer Interest on Digital Camera Products

Sentiment analysis is a process that aims to determine the contents of a text-form dataset (documents, sentences, paragraphs, etc.) positive, negative or neutral. In recent years many sentimental analysis and opinion mining applications have been developed to analyze opinions, feelings and attitudes about products, brands, and news, and others. In general, the review given has a rating that can be determined by the customer itself. However, the ratings provided do not necessarily indicate the content of the reviews submitted. Readers reviews in general cannot judge a review only through rating. The assessment of a review should be done by reading the entire contents of the review. In some studies in the field of sentiment analysis, feature selection is proven to make the classifier more efficient and effective by reducing the amount of data being analyzed, as well as identifying the appropriate features to be considered in the learning process. This study aims to add a feature selection method with wrapper, in this case Genetic Algorithm for sentiment analysis on camera review using Support Vector Machine. Before using Genetic Algortihm, the accuracy of Support Vector Machine was 58.67% and the AUC value was 0.920. After the addition of Genetic Algortihm, the accuracy increased 29.33% to reach 88% and AUC 0.922.